[PATCH] ext3 indexed directory support

Daniel Phillips' indexed directory support.  Ported from ext2 by
Christopher Li.  Contributions from Andreas Dilger, Stephen Tweedie,
lots from Ted.

It requires e2fsprogs-1.29; I've updated the Changes file to reflect
that.

Documentation/Changes

@@ -53,7 +53,7 @@ o  Gnu make               3.78                    # make --version
 o  binutils               2.9.5.0.25              # ld -v
 o  util-linux             2.10o                   # fdformat --version
 o  modutils               2.4.2                   # insmod -V
-o  e2fsprogs              1.25                    # tune2fs
+o  e2fsprogs              1.29                    # tune2fs
 o  jfsutils               1.0.14                  # fsck.jfs -V
 o  reiserfsprogs          3.6.3                   # reiserfsck -V 2>&1|grep reiserfsprogs
 o  xfsprogs               2.1.0                   # xfs_db -V

fs/ext3/Makefile

@@ -5,6 +5,6 @@
 obj-$(CONFIG_EXT3_FS) += ext3.o
 
 ext3-objs    := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \
-		ioctl.o namei.o super.o symlink.o
+		ioctl.o namei.o super.o symlink.o hash.o
 
 include $(TOPDIR)/Rules.make

fs/ext3/dir.c

@@ -23,12 +23,16 @@
 #include <linux/ext3_fs.h>
 #include <linux/buffer_head.h>
 #include <linux/smp_lock.h>
+#include <linux/slab.h>
+#include <linux/rbtree.h>
 
 static unsigned char ext3_filetype_table[] = {
 	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
 };
 
 static int ext3_readdir(struct file *, void *, filldir_t);
+static int ext3_dx_readdir(struct file * filp,
+			   void * dirent, filldir_t filldir);
 
 struct file_operations ext3_dir_operations = {
 	.read		= generic_read_dir,
@@ -37,6 +41,17 @@ struct file_operations ext3_dir_operations = {
 	.fsync		= ext3_sync_file,		/* BKL held */
 };
 
+
+static unsigned char get_dtype(struct super_block *sb, int filetype)
+{
+	if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) ||
+	    (filetype >= EXT3_FT_MAX))
+		return DT_UNKNOWN;
+
+	return (ext3_filetype_table[filetype]);
+}
+			       
+
 int ext3_check_dir_entry (const char * function, struct inode * dir,
 			  struct ext3_dir_entry_2 * de,
 			  struct buffer_head * bh,
@@ -83,6 +98,16 @@ static int ext3_readdir(struct file * filp,
 
 	sb = inode->i_sb;
 
+	if (is_dx(inode)) {
+		err = ext3_dx_readdir(filp, dirent, filldir);
+		if (err != ERR_BAD_DX_DIR)
+			return err;
+		/*
+		 * We don't set the inode dirty flag since it's not
+		 * critical that it get flushed back to the disk.
+		 */
+		EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL;
+	}
 	stored = 0;
 	bh = NULL;
 	offset = filp->f_pos & (sb->s_blocksize - 1);
@@ -167,18 +192,12 @@ static int ext3_readdir(struct file * filp,
 				 * during the copy operation.
 				 */
 				unsigned long version = filp->f_version;
-				unsigned char d_type = DT_UNKNOWN;
 
-				if (EXT3_HAS_INCOMPAT_FEATURE(sb,
-						EXT3_FEATURE_INCOMPAT_FILETYPE)
-						&& de->file_type < EXT3_FT_MAX)
-					d_type =
-					  ext3_filetype_table[de->file_type];
 				error = filldir(dirent, de->name,
 						de->name_len,
 						filp->f_pos,
 						le32_to_cpu(de->inode),
-						d_type);
+						get_dtype(sb, de->file_type));
 				if (error)
 					break;
 				if (version != filp->f_version)
@@ -194,3 +213,269 @@ static int ext3_readdir(struct file * filp,
 	unlock_kernel();
 	return 0;
 }
+
+#ifdef CONFIG_EXT3_INDEX
+/*
+ * These functions convert from the major/minor hash to an f_pos
+ * value.
+ * 
+ * Currently we only use major hash numer.  This is unfortunate, but
+ * on 32-bit machines, the same VFS interface is used for lseek and
+ * llseek, so if we use the 64 bit offset, then the 32-bit versions of
+ * lseek/telldir/seekdir will blow out spectacularly, and from within
+ * the ext2 low-level routine, we don't know if we're being called by
+ * a 64-bit version of the system call or the 32-bit version of the
+ * system call.  Worse yet, NFSv2 only allows for a 32-bit readdir
+ * cookie.  Sigh.
+ */
+#define hash2pos(major, minor)	(major >> 1)
+#define pos2maj_hash(pos)	((pos << 1) & 0xffffffff)
+#define pos2min_hash(pos)	(0)
+
+/*
+ * This structure holds the nodes of the red-black tree used to store
+ * the directory entry in hash order.
+ */
+struct fname {
+	__u32		hash;
+	__u32		minor_hash;
+	struct rb_node	rb_hash; 
+	struct fname	*next;
+	__u32		inode;
+	__u8		name_len;
+	__u8		file_type;
+	char		name[0];
+};
+
+/*
+ * This functoin implements a non-recursive way of freeing all of the
+ * nodes in the red-black tree.
+ */
+static void free_rb_tree_fname(struct rb_root *root)
+{
+	struct rb_node	*n = root->rb_node;
+	struct rb_node	*parent;
+	struct fname	*fname;
+
+	while (n) {
+		/* Do the node's children first */
+		if ((n)->rb_left) {
+			n = n->rb_left;
+			continue;
+		}
+		if (n->rb_right) {
+			n = n->rb_right;
+			continue;
+		}
+		/*
+		 * The node has no children; free it, and then zero
+		 * out parent's link to it.  Finally go to the
+		 * beginning of the loop and try to free the parent
+		 * node.
+		 */
+		parent = n->rb_parent;
+		fname = rb_entry(n, struct fname, rb_hash);
+		kfree(fname);
+		if (!parent)
+			root->rb_node = 0;
+		else if (parent->rb_left == n)
+			parent->rb_left = 0;
+		else if (parent->rb_right == n)
+			parent->rb_right = 0;
+		n = parent;
+	}
+	root->rb_node = 0;
+}
+
+
+struct dir_private_info *create_dir_info(loff_t pos)
+{
+	struct dir_private_info *p;
+
+	p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL);
+	if (!p)
+		return NULL;
+	p->root.rb_node = 0;
+	p->curr_node = 0;
+	p->extra_fname = 0;
+	p->last_pos = 0;
+	p->curr_hash = pos2maj_hash(pos);
+	p->curr_minor_hash = pos2min_hash(pos);
+	p->next_hash = 0;
+	return p;
+}
+
+void ext3_htree_free_dir_info(struct dir_private_info *p)
+{
+	free_rb_tree_fname(&p->root);
+	kfree(p);
+}
+		
+/*
+ * Given a directory entry, enter it into the fname rb tree.
+ */
+void ext3_htree_store_dirent(struct file *dir_file, __u32 hash,
+			     __u32 minor_hash,
+			     struct ext3_dir_entry_2 *dirent)
+{
+	struct rb_node **p, *parent = NULL;
+	struct fname * fname, *new_fn;
+	struct dir_private_info *info;
+	int len;
+
+	info = (struct dir_private_info *) dir_file->private_data;
+	p = &info->root.rb_node;
+
+	/* Create and allocate the fname structure */
+	len = sizeof(struct fname) + dirent->name_len + 1;
+	new_fn = kmalloc(len, GFP_KERNEL);
+	memset(new_fn, 0, len);
+	new_fn->hash = hash;
+	new_fn->minor_hash = minor_hash;
+	new_fn->inode = le32_to_cpu(dirent->inode);
+	new_fn->name_len = dirent->name_len;
+	new_fn->file_type = dirent->file_type;
+	memcpy(new_fn->name, dirent->name, dirent->name_len);
+	new_fn->name[dirent->name_len] = 0;
+	
+	while (*p) {
+		parent = *p;
+		fname = rb_entry(parent, struct fname, rb_hash);
+
+		/*
+		 * If the hash and minor hash match up, then we put
+		 * them on a linked list.  This rarely happens...
+		 */
+		if ((new_fn->hash == fname->hash) &&
+		    (new_fn->minor_hash == fname->minor_hash)) {
+			new_fn->next = fname->next;
+			fname->next = new_fn;
+			return;
+		}
+			
+		if (new_fn->hash < fname->hash)
+			p = &(*p)->rb_left;
+		else if (new_fn->hash > fname->hash)
+			p = &(*p)->rb_right;
+		else if (new_fn->minor_hash < fname->minor_hash)
+			p = &(*p)->rb_left;
+		else /* if (new_fn->minor_hash > fname->minor_hash) */
+			p = &(*p)->rb_right;
+	}
+
+	rb_link_node(&new_fn->rb_hash, parent, p);
+	rb_insert_color(&new_fn->rb_hash, &info->root);
+}
+
+
+
+/*
+ * This is a helper function for ext3_dx_readdir.  It calls filldir
+ * for all entres on the fname linked list.  (Normally there is only
+ * one entry on the linked list, unless there are 62 bit hash collisions.)
+ */
+static int call_filldir(struct file * filp, void * dirent,
+			filldir_t filldir, struct fname *fname)
+{
+	struct dir_private_info *info = filp->private_data;
+	loff_t	curr_pos;
+	struct inode *inode = filp->f_dentry->d_inode;
+	struct super_block * sb;
+	int error;
+
+	sb = inode->i_sb;
+	
+	if (!fname) {
+		printk("call_filldir: called with null fname?!?\n");
+		return 0;
+	}
+	curr_pos = hash2pos(fname->hash, fname->minor_hash);
+	while (fname) {
+		error = filldir(dirent, fname->name,
+				fname->name_len, curr_pos, 
+				fname->inode,
+				get_dtype(sb, fname->file_type));
+		if (error) {
+			filp->f_pos = curr_pos;
+			info->extra_fname = fname->next;
+			return error;
+		}
+		fname = fname->next;
+	}
+	return 0;
+}
+
+static int ext3_dx_readdir(struct file * filp,
+			 void * dirent, filldir_t filldir)
+{
+	struct dir_private_info *info = filp->private_data;
+	struct inode *inode = filp->f_dentry->d_inode;
+	struct fname *fname;
+	int	ret;
+
+	if (!info) {
+		info = create_dir_info(filp->f_pos);
+		if (!info)
+			return -ENOMEM;
+		filp->private_data = info;
+	}
+
+	/* Some one has messed with f_pos; reset the world */
+	if (info->last_pos != filp->f_pos) {
+		free_rb_tree_fname(&info->root);
+		info->curr_node = 0;
+		info->extra_fname = 0;
+		info->curr_hash = pos2maj_hash(filp->f_pos);
+		info->curr_minor_hash = pos2min_hash(filp->f_pos);
+	}
+
+	/*
+	 * If there are any leftover names on the hash collision
+	 * chain, return them first.
+	 */
+	if (info->extra_fname &&
+	    call_filldir(filp, dirent, filldir, info->extra_fname))
+		goto finished;
+
+	if (!info->curr_node)
+		info->curr_node = rb_first(&info->root);
+
+	while (1) {
+		/*
+		 * Fill the rbtree if we have no more entries,
+		 * or the inode has changed since we last read in the
+		 * cached entries. 
+		 */
+		if ((!info->curr_node) ||
+		    (filp->f_version != inode->i_version)) {
+			info->curr_node = 0;
+			free_rb_tree_fname(&info->root);
+			filp->f_version = inode->i_version;
+			ret = ext3_htree_fill_tree(filp, info->curr_hash,
+						   info->curr_minor_hash,
+						   &info->next_hash);
+			if (ret < 0)
+				return ret;
+			if (ret == 0)
+				break;
+			info->curr_node = rb_first(&info->root);
+		}
+
+		fname = rb_entry(info->curr_node, struct fname, rb_hash);
+		info->curr_hash = fname->hash;
+		info->curr_minor_hash = fname->minor_hash;
+		if (call_filldir(filp, dirent, filldir, fname))
+			break;
+
+		info->curr_node = rb_next(info->curr_node);
+		if (!info->curr_node) {
+			info->curr_hash = info->next_hash;
+			info->curr_minor_hash = 0;
+		}
+	}
+finished:
+	info->last_pos = filp->f_pos;
+	UPDATE_ATIME(inode);
+	return 0;
+}
+#endif

fs/ext3/file.c

@@ -34,6 +34,9 @@ static int ext3_release_file (struct inode * inode, struct file * filp)
 {
 	if (filp->f_mode & FMODE_WRITE)
 		ext3_discard_prealloc (inode);
+	if (is_dx(inode) && filp->private_data)
+		ext3_htree_free_dir_info(filp->private_data);
+
 	return 0;
 }
 

fs/ext3/hash.c

+/*
+ *  linux/fs/ext3/hash.c
+ *
+ * Copyright (C) 2002 by Theodore Ts'o
+ *
+ * This file is released under the GPL v2.
+ * 
+ * This file may be redistributed under the terms of the GNU Public
+ * License.
+ */
+
+#include <linux/fs.h>
+#include <linux/jbd.h>
+#include <linux/sched.h>
+#include <linux/ext3_fs.h>
+
+#define DELTA 0x9E3779B9
+
+static void TEA_transform(__u32 buf[4], __u32 const in[])
+{
+	__u32	sum = 0;
+	__u32	b0 = buf[0], b1 = buf[1];
+	__u32	a = in[0], b = in[1], c = in[2], d = in[3];
+	int	n = 16;
+
+	do {							
+		sum += DELTA;					
+		b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b);	
+		b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d);	
+	} while(--n);
+
+	buf[0] += b0;
+	buf[1] += b1;
+}
+
+/* F, G and H are basic MD4 functions: selection, majority, parity */
+#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z))))
+#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z)))
+#define H(x, y, z) ((x) ^ (y) ^ (z))
+
+/*
+ * The generic round function.  The application is so specific that
+ * we don't bother protecting all the arguments with parens, as is generally
+ * good macro practice, in favor of extra legibility.
+ * Rotation is separate from addition to prevent recomputation
+ */
+#define ROUND(f, a, b, c, d, x, s)	\
+	(a += f(b, c, d) + x, a = (a << s) | (a >> (32-s)))
+#define K1 0
+#define K2 013240474631UL
+#define K3 015666365641UL
+
+/*
+ * Basic cut-down MD4 transform.  Returns only 32 bits of result.
+ */
+static void halfMD4Transform (__u32 buf[4], __u32 const in[])
+{
+	__u32	a = buf[0], b = buf[1], c = buf[2], d = buf[3];
+
+	/* Round 1 */
+	ROUND(F, a, b, c, d, in[0] + K1,  3);
+	ROUND(F, d, a, b, c, in[1] + K1,  7);
+	ROUND(F, c, d, a, b, in[2] + K1, 11);
+	ROUND(F, b, c, d, a, in[3] + K1, 19);
+	ROUND(F, a, b, c, d, in[4] + K1,  3);
+	ROUND(F, d, a, b, c, in[5] + K1,  7);
+	ROUND(F, c, d, a, b, in[6] + K1, 11);
+	ROUND(F, b, c, d, a, in[7] + K1, 19);
+
+	/* Round 2 */
+	ROUND(G, a, b, c, d, in[1] + K2,  3);
+	ROUND(G, d, a, b, c, in[3] + K2,  5);
+	ROUND(G, c, d, a, b, in[5] + K2,  9);
+	ROUND(G, b, c, d, a, in[7] + K2, 13);
+	ROUND(G, a, b, c, d, in[0] + K2,  3);
+	ROUND(G, d, a, b, c, in[2] + K2,  5);
+	ROUND(G, c, d, a, b, in[4] + K2,  9);
+	ROUND(G, b, c, d, a, in[6] + K2, 13);
+
+	/* Round 3 */
+	ROUND(H, a, b, c, d, in[3] + K3,  3);
+	ROUND(H, d, a, b, c, in[7] + K3,  9);
+	ROUND(H, c, d, a, b, in[2] + K3, 11);
+	ROUND(H, b, c, d, a, in[6] + K3, 15);
+	ROUND(H, a, b, c, d, in[1] + K3,  3);
+	ROUND(H, d, a, b, c, in[5] + K3,  9);
+	ROUND(H, c, d, a, b, in[0] + K3, 11);
+	ROUND(H, b, c, d, a, in[4] + K3, 15);
+
+	buf[0] += a;
+	buf[1] += b;
+	buf[2] += c;
+	buf[3] += d;
+}
+
+#undef ROUND
+#undef F
+#undef G
+#undef H
+#undef K1
+#undef K2
+#undef K3
+
+/* The old legacy hash */
+static __u32 dx_hack_hash (const char *name, int len)
+{
+	__u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9;
+	while (len--) {
+		__u32 hash = hash1 + (hash0 ^ (*name++ * 7152373));
+		
+		if (hash & 0x80000000) hash -= 0x7fffffff;
+		hash1 = hash0;
+		hash0 = hash;
+	}
+	return (hash0 << 1);
+}
+
+static void str2hashbuf(const char *msg, int len, __u32 *buf, int num)
+{
+	__u32	pad, val;
+	int	i;
+
+	pad = (__u32)len | ((__u32)len << 8);
+	pad |= pad << 16;
+
+	val = pad;
+	if (len > num*4)
+		len = num * 4;
+	for (i=0; i < len; i++) {
+		if ((i % 4) == 0)
+			val = pad;
+		val = msg[i] + (val << 8);
+		if ((i % 4) == 3) {
+			*buf++ = val;
+			val = pad;
+			num--;
+		}
+	}
+	if (--num >= 0)
+		*buf++ = val;
+	while (--num >= 0)
+		*buf++ = pad;
+}
+
+/*
+ * Returns the hash of a filename.  If len is 0 and name is NULL, then
+ * this function can be used to test whether or not a hash version is
+ * supported.
+ * 
+ * The seed is an 4 longword (32 bits) "secret" which can be used to
+ * uniquify a hash.  If the seed is all zero's, then some default seed
+ * may be used.
+ * 
+ * A particular hash version specifies whether or not the seed is
+ * represented, and whether or not the returned hash is 32 bits or 64
+ * bits.  32 bit hashes will return 0 for the minor hash.
+ */
+int ext3fs_dirhash(const char *name, int len, struct dx_hash_info *hinfo)
+{
+	__u32	hash;
+	__u32	minor_hash = 0;
+	const char	*p;
+	int		i;
+	__u32 		in[8], buf[4];
+
+	/* Initialize the default seed for the hash checksum functions */
+	buf[0] = 0x67452301;
+	buf[1] = 0xefcdab89;
+	buf[2] = 0x98badcfe;
+	buf[3] = 0x10325476;
+
+	/* Check to see if the seed is all zero's */
+	if (hinfo->seed) {
+		for (i=0; i < 4; i++) {
+			if (hinfo->seed[i])
+				break;
+		}
+		if (i < 4)
+			memcpy(buf, hinfo->seed, sizeof(buf));
+	}
+		
+	switch (hinfo->hash_version) {
+	case DX_HASH_LEGACY:
+		hash = dx_hack_hash(name, len);
+		break;
+	case DX_HASH_HALF_MD4:
+		p = name;
+		while (len > 0) {
+			str2hashbuf(p, len, in, 8);
+			halfMD4Transform(buf, in);
+			len -= 32;
+			p += 32;
+		}
+		minor_hash = buf[2];
+		hash = buf[1];
+		break;
+	case DX_HASH_TEA:
+		p = name;
+		while (len > 0) {
+			str2hashbuf(p, len, in, 4);
+			TEA_transform(buf, in);
+			len -= 16;
+			p += 16;
+		}
+		hash = buf[0];
+		minor_hash = buf[1];
+		break;
+	default:
+		hinfo->hash = 0;
+		return -1;
+	}
+	hinfo->hash = hash & ~1;
+	hinfo->minor_hash = minor_hash;
+	return 0;
+}

fs/ext3/super.c

@@ -762,6 +762,7 @@ static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
 	es->s_mtime = cpu_to_le32(CURRENT_TIME);
 	ext3_update_dynamic_rev(sb);
 	EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
+
 	ext3_commit_super (sb, es, 1);
 	if (test_opt (sb, DEBUG))
 		printk (KERN_INFO
@@ -772,6 +773,7 @@ static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
 			EXT3_BLOCKS_PER_GROUP(sb),
 			EXT3_INODES_PER_GROUP(sb),
 			sbi->s_mount_opt);
+
 	printk(KERN_INFO "EXT3 FS " EXT3FS_VERSION ", " EXT3FS_DATE " on %s, ",
 				sb->s_id);
 	if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
@@ -946,6 +948,7 @@ static loff_t ext3_max_size(int bits)
 	return res;
 }
 
+
 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 {
 	struct buffer_head * bh;
@@ -1115,6 +1118,9 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 	sbi->s_mount_state = le16_to_cpu(es->s_state);
 	sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
 	sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
+	for (i=0; i < 4; i++)
+		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
+	sbi->s_def_hash_version = es->s_def_hash_version;
 
 	if (sbi->s_blocks_per_group > blocksize * 8) {
 		printk (KERN_ERR

include/linux/ext3_fs.h

@@ -40,6 +40,11 @@
 #define EXT3FS_DATE		"02 Dec 2001"
 #define EXT3FS_VERSION		"2.4-0.9.16"
 
+/*
+ * Always enable hashed directories
+ */
+#define CONFIG_EXT3_INDEX
+
 /*
  * Debug code
  */
@@ -440,8 +445,11 @@ struct ext3_super_block {
 /*E0*/	__u32	s_journal_inum;		/* inode number of journal file */
 	__u32	s_journal_dev;		/* device number of journal file */
 	__u32	s_last_orphan;		/* start of list of inodes to delete */
-
-/*EC*/	__u32	s_reserved[197];	/* Padding to the end of the block */
+	__u32	s_hash_seed[4];		/* HTREE hash seed */
+	__u8	s_def_hash_version;	/* Default hash version to use */
+	__u8	s_reserved_char_pad;
+	__u16	s_reserved_word_pad;
+	__u32	s_reserved[192];	/* Padding to the end of the block */
 };
 
 #ifdef __KERNEL__
@@ -584,8 +592,45 @@ struct ext3_dir_entry_2 {
 #define EXT3_DIR_ROUND			(EXT3_DIR_PAD - 1)
 #define EXT3_DIR_REC_LEN(name_len)	(((name_len) + 8 + EXT3_DIR_ROUND) & \
 					 ~EXT3_DIR_ROUND)
+/*
+ * Hash Tree Directory indexing
+ * (c) Daniel Phillips, 2001
+ */
+
+#ifdef CONFIG_EXT3_INDEX
+  #define is_dx(dir) (EXT3_HAS_COMPAT_FEATURE(dir->i_sb, \
+					      EXT3_FEATURE_COMPAT_DIR_INDEX) && \
+		      (EXT3_I(dir)->i_flags & EXT3_INDEX_FL))
+#define EXT3_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT3_LINK_MAX)
+#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
+#else
+  #define is_dx(dir) 0
+#define EXT3_DIR_LINK_MAX(dir) ((dir)->i_nlink >= EXT3_LINK_MAX)
+#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2)
+#endif
+
+/* Legal values for the dx_root hash_version field: */
+
+#define DX_HASH_LEGACY		0
+#define DX_HASH_HALF_MD4	1
+#define DX_HASH_TEA		2
+
+/* hash info structure used by the directory hash */
+struct dx_hash_info
+{
+	u32		hash;
+	u32		minor_hash;
+	int		hash_version;
+	u32		*seed;
+};
 
 #ifdef __KERNEL__
+/*
+ * Control parameters used by ext3_htree_next_block
+ */
+#define HASH_NB_ALWAYS		1
+
+
 /*
  * Describe an inode's exact location on disk and in memory
  */
@@ -596,6 +641,27 @@ struct ext3_iloc
 	unsigned long block_group;
 };
 
+
+/*
+ * This structure is stuffed into the struct file's private_data field
+ * for directories.  It is where we put information so that we can do
+ * readdir operations in hash tree order.
+ */
+struct dir_private_info {
+	struct rb_root	root;
+	struct rb_node	*curr_node;
+	struct fname	*extra_fname;
+	loff_t		last_pos;
+	__u32		curr_hash;
+	__u32		curr_minor_hash;
+	__u32		next_hash;
+};
+
+/*
+ * Special error return code only used by dx_probe() and its callers.
+ */
+#define ERR_BAD_DX_DIR	-75000
+
 /*
  * Function prototypes
  */
@@ -623,11 +689,20 @@ extern struct ext3_group_desc * ext3_get_group_desc(struct super_block * sb,
 
 /* dir.c */
 extern int ext3_check_dir_entry(const char *, struct inode *,
-				struct ext3_dir_entry_2 *, struct buffer_head *,
-				unsigned long);
+				struct ext3_dir_entry_2 *,
+				struct buffer_head *, unsigned long);
+extern void ext3_htree_store_dirent(struct file *dir_file, __u32 hash,
+				    __u32 minor_hash,
+				    struct ext3_dir_entry_2 *dirent);
+extern void ext3_htree_free_dir_info(struct dir_private_info *p);
+
 /* fsync.c */
 extern int ext3_sync_file (struct file *, struct dentry *, int);
 
+/* hash.c */
+extern int ext3fs_dirhash(const char *name, int len, struct
+			  dx_hash_info *hinfo);
+
 /* ialloc.c */
 extern struct inode * ext3_new_inode (handle_t *, struct inode *, int);
 extern void ext3_free_inode (handle_t *, struct inode *);
@@ -636,6 +711,7 @@ extern unsigned long ext3_count_free_inodes (struct super_block *);
 extern void ext3_check_inodes_bitmap (struct super_block *);
 extern unsigned long ext3_count_free (struct buffer_head *, unsigned);
 
+
 /* inode.c */
 extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *);
 extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *);
@@ -659,6 +735,8 @@ extern int ext3_ioctl (struct inode *, struct file *, unsigned int,
 /* namei.c */
 extern int ext3_orphan_add(handle_t *, struct inode *);
 extern int ext3_orphan_del(handle_t *, struct inode *);
+extern int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash,
+				__u32 start_minor_hash, __u32 *next_hash);
 
 /* super.c */
 extern void ext3_error (struct super_block *, const char *, const char *, ...)

include/linux/ext3_fs_sb.h

@@ -48,6 +48,8 @@ struct ext3_sb_info {
 	int s_inode_size;
 	int s_first_ino;
 	u32 s_next_generation;
+	u32 s_hash_seed[4];
+	int s_def_hash_version;
 
 	/* Journaling */
 	struct inode * s_journal_inode;

include/linux/ext3_jbd.h

@@ -63,6 +63,8 @@ extern int ext3_writepage_trans_blocks(struct inode *inode);
 
 #define EXT3_RESERVE_TRANS_BLOCKS	12
 
+#define EXT3_INDEX_EXTRA_TRANS_BLOCKS	8
+
 int
 ext3_mark_iloc_dirty(handle_t *handle, 
 		     struct inode *inode,

include/linux/rbtree.h

@@ -122,6 +122,7 @@ extern void rb_erase(struct rb_node *, struct rb_root *);
 /* Find logical next and previous nodes in a tree */
 extern struct rb_node *rb_next(struct rb_node *);
 extern struct rb_node *rb_prev(struct rb_node *);
+extern struct rb_node *rb_first(struct rb_root *);
 
 /* Fast replacement of a single node without remove/rebalance/add/rebalance */
 extern void rb_replace_node(struct rb_node *victim, struct rb_node *new, 

lib/rbtree.c

@@ -296,6 +296,22 @@ void rb_erase(struct rb_node *node, struct rb_root *root)
 }
 EXPORT_SYMBOL(rb_erase);
 
+/*
+ * This function returns the first node (in sort order) of the tree.
+ */
+struct rb_node *rb_first(struct rb_root *root)
+{
+	struct rb_node	*n;
+
+	n = root->rb_node;
+	if (!n)
+		return 0;
+	while (n->rb_left)
+		n = n->rb_left;
+	return n;
+}
+EXPORT_SYMBOL(rb_first);
+
 struct rb_node *rb_next(struct rb_node *node)
 {
 	/* If we have a right-hand child, go down and then left as far